The Drosophila segment polarity gene patched is involved in a position-signalling mechanism in imaginal discs

The body of the Drosophila adult is derived from a series of embryonic primordia (imaginal discs) which undergo extensive rounds of cell division during the larval stages of development (Auerbach, 1936; Nothiger, 1972). Although these cells receive instructions about their segmental identity during embryogenesis (Akam, 1987; Peifer et al. 1987), clonal analysis has shown that their early lineage, with the exception of the anterior-posterior compartmentalisation event (Garcia-Bellido et al. 1973), is largely indeterminate (Garcia-Bellido and Merriam, 1971; Bryant, 1970). The information required for the patterning of cells in different structures is thus an emergent property, which presumably depends upon communication between cells as they proliferate. A large body of data shows that imaginal disc cells respond to changes in the identity of neighbouring cells whether these alterations are made by surgical or genetic changes (reviewed in Whittle, 1990). When normally distant parts of a mature wing disc are juxtaposed and allowed to proliferate, the intervening pattern of cell types is regenerated by a process known as intercalation (Haynie and Bryant, 1976). In some genetic mosaics, wild-type cells respond to neighbours of changed genotype by differentiating cell types in changed patterns (Stem, 1956; Simpson and Schneiderman, 1975; Mohler, 1988; Santamaria et al. 1989) or polarities (Gubb and Garcia-Bellido, 1982; Vinson and Adler, 1987). In contrast to the early events of embryonic patterning (Ingham, 1988), virtually nothing is known about the molecular basis of this process.

Amongst the many mutations that disrupt adult morphogenesis are viable alleles of some of the segment polarity genes, a class originally defined by embryonic lethal mutations (Nüsslein-Volhard and Weischaus, 1980; Nüsslein-Volhard et al. 1984). An increasing body of evidence has implicated these genes in cell interactions (Martinez-Arias et al. 1988; DiNardo et al. 1988; Mohler, 1988). Here we show that mutations of the segment polarity gene patched (ptc) disrupt cell patterning in the cuticle of the adult fly. Using genetic mosaics of patched we find a requirement for this gene in cells in specific regions of the wing, which is not rescued by neighbouring cells; moreover in some locations the effects of patched mutations extend beyond the limit of the mosaic, suggesting a role for the gene in a signal relay mechanism. We describe the pattern of patched transcription in the developing imaginal discs in relation to the expression of certain other genes and compare this to the requirements revealed by the mosaic analysis and the phenotypes of the viable alleles of patched.

14 patched alleles were generously supplied by C. Nüsslein-Volhard. ptc^G12, ptc^G13 and ptc^G20 were recovered in an F, screen at the University of Sussex on a cinnabar, brown, speck chromosome following 400 Grays irradiation from a °°Co source. Eight alleles were recovered in a similar F₁ screen following a dysgenic cross between the Harwich P-element containing stock and the same M cytotype stock used in the irradiation screen. The mutation tufted (Sturtevant, 1948) and T(2:3)dp, which behaves as if it carries an allele of tufted, were obtained from stock centres, ptc^S2 was found as an allele of tufted during a gamma-irradiation mutagenesis screen in this laboratory (Whittle, 1980). Ptc^RX67 was kindly supplied by P. Simpson (see Table 1 for a description and source of ptc alleles), the engrailed-lacZ reporter genotype by C. Hama and T. Kornberg and the neuralized–lacZ reporter by D. Clements and J. Merriam.

Mosaics were induced by 100 Grays gamma irradiation from a ⁶⁰Co source in straw pawn ptc^x/ M(2)cft^33a, stw pwn ptc^x/ M(2)S1 and in stw pwn ptc^x/shavenoid genotypes, ptc^x representing the patched allele under examination. Wings were examined under phase-contrast illumination for the phenotypes of the cell marker mutations pawn (pwn) and shavenoid (sha) (Fig. 2).

Larvae and prepupae were collected from the media and walls of a yeasted cornmeal-agar milk bottle, washed and dissected in phosphate-buffered saline (PBS), pH 7.2. The anterior halves of animals were separated, inverted and stored on ice for less than 30 min; fixed for 15 min on ice in 4 % paraformaldehyde in PBS and washed three times for 1 min in diethylpyr-ocarbonate-treated PBS plus 0.1% Tween 20 (PBT). The prehybridization treatment and hybridization conditions are based on the non-radioactive whole mount in situ hybridization protocol for Drosophila embryos of Tautz and Pfeifle (1989). The inverted heads were digested for 3 min at 22°C in 50 μg ml^-1 proteinase K in PBT and washed twice for 2 min in 2 mg ml^-1 glycine in PBT. After two 1 min washes in PBT, the heads were fixed for a further 20 min in 4% paraformaldehyde in PBS at 22°C, then washed for 5 min in 2 mg ml^-1 glycine in PBT, twice for 5 min in PBT, 5 min in PBT:hybridization buffer (1:1), and 3 –5 h in hybridization buffer (50% formamide, 5 ×SSC, 100 μgml^-1 denatured herring sperm DNA and 0.1 % Tween 20) at 45°C. The heads were hybridized overnight at 45 °C in hybridization buffer containing 1 μgml^-1 denatured digoxigenin-dUTP-labelled DNA. Heads were washed for 20 min at 22 °C in the hybridization solution then in a series of hybridization solution: PBT dilutions (4:1, 3:2, 2:3, 1:4) for 20 min each and finally in PBT twice for 20min. Incubation in polyclonal sheep anti-digoxigenin Fab fragments conjugated to alkaline phosphatase (Boehringer Mannheim) at 0.375 i.u. ml^-1 in PBT for 1h at 22 °C was followed by four 20 min washes each in PBT and two 5 min washes in AP buffer (100 mM NaCl, 50 mM MgCl₂, 100 mM Tris pH 9.2 and 0.1 % Tween 20). Finally, they were developed in the ‘brown’ alkaline phosphatase substrate (Kit III, SK-5300, Vector Laboratories) or in AP buffer with nitroblue tétrazolium salt (337.5 μg ml^-1) and 5-bromo-4 chloro-3-indolyl phosphate toluidinium salt (175 μgml^-1) for 1 h at 22°C in the dark. The reaction was stopped by transfer to PBT. Individual discs were dissected in PBS, mounted in Hydromount (National Diagnostics) and photographed using Differential Interference Contrast photomicroscopy. For detection of β-galactosidase activity, the material was incubated for 10 min (en–lacZ) or 3h (neu–lacZ) at 37 °C in X-gal staining solution (Simon et al. 1985) after the initial fixation and 1 wash in PBT, then the hybridization protocol was continued. Larvae heterozygous for en–lacZ or neu–lacZ were collected from an outcross to Canton S wild-type flies.

The probe for detection of patched expression was prepared by random primer labelling (Feinberg and Vogelstein, 1983) using digoxigenin-modified dUTP and the patched cDNA plasmid, pC7 .This plasmid was isolated (Phillips, unpublished) from an 8 –12 h embryonic cDNA library generously provided by N. Brown using genomic probes from the patched region (Nakano et al. 1989). The probe was prepared using the Non-radioactive DNA Labelling and Detection Kit (cat. no. 1093 657, Boehringer Mannheim) as described in the instructions, using 100 ng of linear template DNA in a 20 μl reaction.

A new allele of the adult-viable recessive mutation tufted (Sturtevant, 1948) found in this laboratory (Whittle, 1980) is an embryonic recessive-lethal mutation with a segment-polarity larval cuticular phenotype identical to that of patched (Nüsslein-Volhard and Wieschaus, 1980; Nüsslein-Volhard et al. 1984). All 14 extant patched lethal alleles failed to complement the lethality of this new allele of tufted, producing instead embryos arrested with a typical patched larval cuticle; in addition they failed to complement the adult-viable phenotype of tufted (Fig. 1). This new lethal allele was therefore designated ptc^S2 and tufted redesignated ptc^tuf.

We carried out F₁ mutagenesis screens to identify new mutations that failed to complement features of the ptc^tuf phenotype. This procedure yielded 10 new lethal alleles of patched, and another adult-viable allele (Table 1). Among the 10 new lethal alleles, 8 were induced by P-M-dysgenesis, four of these being associated with P insertions at 44D3,4 (Hidalgo, unpublished). One of these, ptc^p78 has been used to clone the patched gene (Nakano et al. 1989).

The combined evidence from the heteroallelic combinations of the 26 lethal and 3 viable alleles shows that the patterning of structures formed by all the imaginal discs and histoblast nests is dependent upon patched activity. Of these 78 heteroallelic combinations, only 5 have a wild-type phenotype (Roberts and Whittle, unpublished). Here we report only upon the effects within the wing disc and its derivatives.

There is considerable phenotypic variability in the effects on the wing both within and between different patched genotypes. Wing phenotypes of ptc^tuf homozygotes range from wild-type to loss of costal bristles and moderate duplication of the base of vein 1 with the adjacent triple row margin. In the most subtle changes seen in the wing blade from this genotype, the base of vein 1 at its junction with the costa may be swollen or bulge out anteriorly. More extensive changes at the anterior margin include the loss of bristles of the medial and distal costa (Fig. 1B), broadening of the region between vein 1 and vein 2 with accompanying change of shape in vein 1 (Fig. 1B), or duplications of the base of vein 1 and associated triple row bristles (not shown). Combinations of ptc^tuf with lethal patched alleles cause more extensive deletions of the costa and duplications of veins 1 and 2. Vein 2 may be duplicated partially or through its entire length and the triple row bristles often show a change in their orientation (or ‘polarity’) between the sites at which the duplicate veins touch the margin (Fig. 1C and E). This local duplication of structures forms part of the most extreme change seen in patched wings in which the costa is also absent and the wing shape is broader, foreshortened and increased in the area occupied by the duplicated vein 2 (Fig. 1C and E). More or less severe changes of this kind are typical of the majority of adult-viable patched genotypes.

A contrasting phenotypic change in vein 2 and upsets involving vein 3 are seen only in genotypes that include the viable allele ptc^G20. In ptc^G20 homozygotes the costal, triple row and double row bristles remain intact and show no polarity reversals but the wing shape and vein pattern are changed. In these wings, vein 2 is partially or completely deleted and veins 1 and 3 are broadened and/or plexate. In combination with lethal patched alleles, ptc^G20 may cause loss of bristles in the costa, but the phenotypes of veins 1, 2 and 3 are not substantially different from those of ptc^G20 homozygotes (Fig. 1D). In contrast to the changes seen in the anterior compartment, no patched genotype has yet been found in which posterior compartment structures are upset.

Although allelism between patched and tufted mutations was established on the basis of the wing blade disturbance, many combinations of patched lethal alleles with ptc^tuf have their more obvious effects in the derivatives of the eye antennal disc, or increase the number and upset the pattern of large bristles on the notum, scutellum and abdominal tergites (Roberts and Whittle, unpublished). The range and type of wing phenotypic changes described, illustrated in Fig. 1, make it clear that there is no simple ‘phenotypic series’ of graded and increasingly severe effects amongst the allelic combinations of patched mutations; indeed some combinations have phenotypically wild-type wings but severe disturbances in other disc derivatives.

To analyse the cellular basis of the patched requirement in wings, we generated mosaic animals bearing clones of cells homozygous for patched lethal mutations (which we shall call ptc clones). The behaviour of ptc clones induced during wing disc growth by mitotic recombination could be divided into three categories, showing an important dependence upon the position of the clone. In region 1 of the wing blade (see Figs 1A and 3A for the limits of these regions), the recovery of ptc clones was significantly reduced, whilst in region 2 ptc clones affected vein pattern both autonomously and in neighbouring wild-type cells, and, in region 3, including the entire posterior compartment, there were no phenotypic consequences of the presence of the ptc clone.

There is a significant shortfall in ptc clones recovered between the anterior wing margin and vein 2 (Fig. 1A and Table 2). Mitotic recombination in larvae of the genotype stw pwn ptc^S2/sha, proximal to straw, should produce twin daughter clones which are either ptc^- or ptc⁺ marked by straw pawn and shavenoid, respectively. The twin/single shavenoid clone ratio between the anterior margin and vein 2 is 0/14 compared to 35/68 in the remainder of the wing (P<0.001, Fisher’s exact 2 × 2 test). Of 98 marked clones in total, only three lay in this area whereas this fraction for multiple wing hairs clones is 35/51 (Whittle, unpublished). Anterior compartment ptc clones are occasionally fragmented (5 cases, Fig. 2D) or considerably smaller than their shavenoid twin, particularly when the shavenoid twin is anterior to the ptc clone (6 cases in 20 sets of twins, Fig. 2B). Two lines of evidence suggest that region 1 extends posterior to vein 2. ptc clones are infrequent along the posterior edge of vein 2 as well as anterior to this vein and never form triple row bristles. One of the surviving ptc clones recovered in region 1 was associated with ectopic venation and we envisage that the essential requirement for patched in region 1 is superimposed on a vein patterning role similar to that in region 2 (see below). In vivo culture of disc primordia from embryos arrested during segmentation from two different hetero-allelic combinations of patched embryonic lethal alleles (ptc^IN/ptc^G12 and ptc^S2/ptc^G12) (Simcox et al. 1989 and Simcox and Whittle, unpublished) provides confirming evidence that patched is required for development of region 1. The cells of these disc primordia proliferate in the abdomen of adult flies and, when transplanted to metamorphosing larvae, form extensive areas of double row bristles and posterior hairs characteristic of regions 2 and 3, but never form costal or triple row bristles (8 embryos tested).

Clones that do form in the anterior compartment never form part of any vein and wild-type cells bordering these clones often form ectopic vein not usually found at these locations (Fig. 2). Ectopic venation was found in 23 of 24 clones between veins 2 and 3, and in 16 of 35 clones between vein 3 and the A/P compartment border (Table 2). 24 clones were totally surrounded by vein tissue made of wild-type cells immediately adjacent to the clones (Fig. 2A,B,D), and all ectopic veins bordered ptc clones, ptc clones associated with extra venation are found in a proximo-distal sector of the anterior compartment on both dorsal and ventral surfaces (Fig. 1A, hatched area), and their borders in this region are smooth and straight (Fig. 2A – E) unlike wild-type clones (Garcia-Bellido and Merriam 1971; Bryant 1970) and ptc clones in the posterior compartment (data not shown). We have seen 7 cases of large ptc clones with straight borders posterior and adjacent to vein 3 that have no ectopic veination (e.g. Fig. 2F). The posterior boundary of region 2 can be defined by a group of 4 clones, including one shown in Fig. 2E, which are surrounded by ectopic vein except at the posterior edge where the clone extends into region 3. The boundary is posterior to the wild-type vein 3 position and anterior to the A/P border. All the data in Table 2 relate to clones of ptc^S2 but similar ectopic venation has been found close to shavenoid clones in the genotype ptc^P76/sha, suggesting that this P-element disrupted allele behaves similarly to ptc^S2.

We have used a non-radioactive method (see Materials and methods) to detect in situ hybridization of a patched cDNA probe to RNA in whole-mount imaginal discs and applied this simultaneously with detection of β - galactosidase activity to discs from stocks carrying lacZ reporter P-element inserts in either the engrailed gene (en – lacZ) or the neuralized gene (neu – lacZ). This novel double-labelling procedure allows us to relate the pattern of patched expression to features of the imaginal disc fate map (Fig. 3A).

In imaginal wing discs from third instar larvae and prepupae, cells in a stripe parallel to the A/P border express patched intensely (Fig. 4A – I)). The stripe follows the border of the posterior compartment as defined by expression of en – lacZ (Fig. 4B – E). No unlabelled cells intervene between patched and en – lacZ expressing cells in the double-labelled preparations. Some cells of the stripe express only patched whilst others express both patched and en – lacZ (Fig. 4C – E). In the wing pouch, posterior to the patched stripe, patched expression decreases abruptly. By contrast, anterior to the stripe, expression decreases over several cell diameters to a lower level which is maintained thoughout the anterior compartment. The contrast between anterior and posterior staining is stronger in prepupal discs (Fig. 4H,I) than in larval discs (Fig. 4A,G).

Interpretation of the relationship of the stripe relative to the compartment border is complicated by the complex three-dimensional structure of the disc which is formed by an invagination of a group of cells that straddle the A/P border in the embryonic epithelium. The invagination flattens and maintains only a narrow attachment to the larval epithelium, called the stalk (Fig. 3A). Therefore the A/P border runs continuously from the stalk, up one side of the flattened disc and down the other side, back to the stalk. One side of the disc is a folded columnar epithelium and on the other the epithelium is stretched to form the squamous cells of the peripodial membrane (Fig. 3B). The wing pouch, the primordium for the adult wing blade, bulges against the peripodial membrane (Fig. 3B). Because the membrane is very thin, the features visible from the peripodial side are primarily those of the columnar wing pouch (Fig. 4A,B,D,F – I). The stripe of patched-expressing cells in the pouch is 2 – 3 cells wide and is interrupted by a gap consisting of 15 – 20 cells expressing patched at a lower level at the intersection of the A/P border with the dorsal-ventral compartment (D/v) border as defined by expression of the neu–lacZ insertion (Fig. 4F – H). β -galactosidase activity is detected in the precursors of bristles and sensilla in the discs of larvae carrying this insertion. In the wing disc, these precursors include two rows of multiply innervated bristles (recurved), which lie on either side of the D/V border on the primordium of the anterior wing margin, as well as campaniform sensilla on veins 1, 2 and 3, and macroachaetes on the costa (Fig. 3A). The stripe of patched expression is separated from the primordia of vein 3 sensilla by 2 – 3 cells (Fig. 4F). In the middle of the disc, the A/P border and the patched stripe descend into and emerge from three major folds (Fig. 4B,C) and cross the stalk near the posterior edge of the disc. At the top of the pouch, the stripe descends into a fold and ascends laterally. Stripes of patched-expressing cells border both sides of the posterior compartment in the stalk (Fig. 4C) as would be expected in a pouch-like invagination. The extension of this border across the upper part of the peripodial side is undetectable probably because these cells are thin, patched is also expressed in the other imaginal discs (Phillips, in preparation).

We have presented genetic evidence that the segmentpolarity gene patched is required in some cells of the wing disc for normal spatial patterning. The expression pattern of patched in the wing disc is complex spatially and temporally. We offer an interpretation of the role of patched that integrates these observations, suggesting it to be a component in a cell-to-cell positionsignalling mechanism which is required in at least two processes in the developing wing: (1) patched is required autonomously for cell viability and/or profiferation in the region adjacent and anterior to vein 2, and (2) patched is required for vein formation throughout the anterior compartment. The role of patched in cell communication is evident in the way that wild-type cells respond to adjacent ptc^- cells by forming ectopic vein.

We distinguish three regions in our analysis of patched (1, 2 and 3, Figs 1A and 3A), which must represent different cellular ‘environments’ with respect to the contribution of patched. In region 1, few ptc clones are recovered and those that do survive are small and/or fragmented and do not form veins. Our twin clone analysis shows that ptc clones are not lost because of migration away from this region or through the failure of flies carrying such clones to eclose, but because the cells of ptc clones fail to proliferate or die before cuticle formation. However, even small clones are infrequent and no disturbance symptomatic of late loss of ptc clones has been seen, favouring the interpretation that the absence of ptc clones is due to a continuous or intermittent requirement for patched throughout disc growth. The finding that costal and triple row bristles are not produced by cultured tissue heteroallelic for patched lethal alleles is further evidence for this requirement.

Cells can survive in region 2 without patched; nevertheless all cells in this region require patched expression for normal development. The consequences of forming patched mosaics are two-fold and differ from those seen in region 1. First, cells require patched before they can participate in forming veins. Second, the removal of patched expression from cells of region 2 causes their immediate neighbours to form veins, so implying a role for patched in signalling between cells in this region. We propose that ptc clones lead to the formation of veins in adjacent cells either because a positive induction signal is generated by default, or because patched is required in a lateral inhibition system for vein spacing in the anterior compartment, ptc clones in the centre of the region competent to form vein 3 (region 2) prompt vein formation on all sides (Fig. 2D,G), whereas clones straddling this region are only bounded by a vein within the region (Fig. 2E,H). Clones in which the majority of cells are outside and posterior to region 2 are subtly different; their anterior borders are atypically straight where the clones have entered region 2 and presumably induced vein 3 (Fig. 2F,I). We suggest that, once cells have become committed to differentiating a vein, a straightening or smoothing process occurs. In wild-type wings, this smoothing process would be undetectable, but in our mosaics, because the clone borders are also the vein borders the adjustment of the presumptive vein cells ‘smooths’ the outline of the marked ptc clone. The approximate position of the veins is established by the end of the larval period (Garcia-BeUido, 1977). However, specification of individual cells to form veins probably occurs during metamorphosis. Expression of patched increases throughout the anterior compartment at this stage, consistent with a role in this process in both regions 1 and 2.

Within region 3 ptc clones have irregular outlines, except where they obey the A/P compartment restriction. Clones in the posterior compartment participate in the formation of veins 4 and 5. Therefore patched is not involved in posterior vein patterning.

The boundaries between regions where ptc clones behave differently (regions 1, 2 3), appear to correspond to the boundaries of ‘zones of positional homology’ defined by the position-dependent response of clones homozygous for shaggy mutations (Ripoll et al. 1988; Simpson et al. 1988). Homozygous shaggy clones differentiate bristles rather than trichomes but the particular bristle morphology varies with position, resembling that of the adjacent marginal bristles defining 5 zones (A – E) in their Fig. 4 (Simpson et al. 1988).

We now consider the phenotypes produced by several adult-viable patched genotypes in the light of our distinction between the patched requirement in regions 1 and 2. The viable alleles, ptc^tuf and ptc^G20, confer distinct spectra of abnormalities. Allelic combinations with ptc^tuf delete parts of the costa and expand the triple row region. The altered patterns of marginal bristles and venation shown in Fig. 1B,C and E, in particular the mirror imaging in Fig. 1E, would be consistent with the consequences of loss of cells with the positional values corresponding to the costa and triple row bristles on the wing disc fate map (Bryant, 1975). However, adult-viable genotypes that include ptc^G20 have a very different wing phenotype (Fig. 1D). Polarity reversals and pattern duplications are very rare but veins are abnormal. Veins 1 and 3 are widened or plexate while vein 2 is often interrupted or deleted. This suggests that ptc^G20 is specifically deficient for the patched activity required in vein regulation while ptc^tuf is deficient for the cell viability function in region 1. If patched is required for both lateral inhibition and specification of vein-forming cells, as the clonal analysis implies, then cells that are hypomorphic for this activity might be competent to form veins but unable to regulate the amount of vein, resulting in broadened veins. Vein 2 might be supressed by cells having either too little or too much activity (see below).

We detect an intense stripe of patched expression that straddles the A/P border in imaginal discs. In contrast, during embryogenesis, patched expression is restricted to cells that do not produce engrailed protein and a regulatory relationship between these genes has been proposed (Martinez-Arias et al. 1988; Dinardo et al. 1988; Hooper and Scott, 1989; Nakano et al. 1989; Hidalgo and Ingham, unpublished). Large ptc clones abutting the A/P border for much of its length fail to show any phenotypic changes (Fig. 2F). We suggest three possible interpretations of this result: (1) if the role of patched in this stripe is not cell-autonomous, patched-expressing cells of the posterior compartment could ‘rescue’ a large anterior ptc clone and vice versa. (2)the expression of patched in this region may have no function, being a relic of the embryonic expression pattern (Nakano et al. 1989; Hooper and Scott, 1989) or (3) the high level of patched expression in this area might have an autonomous role in vein suppression thus defining the posterior boundary of region 2. ptc clones could not be used to detect this function because they are also incompetent to form veins.

The data presented in this paper mean that patched joins the growing collection of genes shown to be important in imaginal disc patterning that are also deployed during embryonic segment patterning, and which are known in that context as segment polarity genes (Nüsslein-Volhard and Wieschaus, 1980; Baker, 1988a, 1988b, Mohler 1988; Perrimon and Mahowald 1987; Busson et al. 1988; Grau and Simpson 1987; Simpson and Grau 1987; Diaz-Benjumea and Garcia-Bellido, 1990). The hydropathy profile of the patched coding sequence suggests that it is a multi-transmembrane domain protein (Nakano et al. 1989; Hooper and Scott 1989). This and the domineering non-autonomy of ptc clones are strongly consistent with our proposed role for patched in cell-to-cell signalling during pattern formation.

The authors record their grateful thanks to Jillion Davey, Nicola Ford and Colin Atherton for considerable assistance in producing this manuscript. This work was supported by an SERC project grant to R.G.P. and J.R.S.W., an SERC graduate studentship to I.J.H.R. and by the ICRF.